training problem

[yanghan0617]

how can i solve this problem.

(pytorch_yh) I:\yh\code\MPR\Deraining>python train.py D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\optim\lr_scheduler.py:131: UserWarning: Detected call of lr_scheduler.step() before optimizer.step(). In PyTorch 1.1.0 and later, you should call them in the opposite order: optimizer.step() before lr_sch eduler.step(). Failure to do this will result in PyTorch skipping the first value of the learning rate schedule. See more details at https://pytorch.org/docs/stable/optim.html#how-to-adjust-learning-rate warnings.warn("Detected call of lr_scheduler.step() before optimizer.step(). " ===> Start Epoch 1 End Epoch 101 ===> Loading datasets 1%|███▏ | 5/347 [00:07<08:45, 1.54s/it] Traceback (most recent call last): File "I:\yh\code\MPR\Deraining\train.py", line 111, in for i, data in enumerate(tqdm(train_loader), 0): File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\tqdm\std.py", line 1182, in iter for obj in iterable: File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data\dataloader.py", line 652, in next data = self._next_data() File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data\dataloader.py", line 692, in _next_data data = self._dataset_fetcher.fetch(index) # may raise StopIteration File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data_utils\fetch.py", line 52, in fetch return self.collate_fn(data) File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data_utils\collate.py", line 175, in default_collate return [default_collate(samples) for samples in transposed] # Backwards compatibility. File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data_utils\collate.py", line 175, in return [default_collate(samples) for samples in transposed] # Backwards compatibility. File "D:\Anaconda\envs\pytorch_yh\lib\site-packages\torch\utils\data_utils\collate.py", line 141, in default_collate return torch.stack(batch, 0, out=out) RuntimeError: stack expects each tensor to be equal size, but got [3, 244, 256] at entry 0 and [3, 104, 256] at entry 1

[Feecuin]

Hello, have you encountered this problem before?

[Feecuin]

你好，你后面可以训了吗？你的损失是否会下降，我的降到18就不会动了，PSNR也上不去

[Feecuin]

老哥，那我这个是啥问题？我按照readme来，只是改了120行和121行，因为报错说不是一个张量，后面改成了这样 可以PSNR上不去，而且训练的很快，我数据集有2000张图的，我怀疑是不是没有正确读图？

[Feecuin]

我也是像你这样改的，我一开始也是不能训练，看的GitHub改的，和你这个一样，然后就可以了。你用的啥机器跑的，用的哪个数据集。 @. … ---- Replied Message ---- From @.> Date 4/1/2024 14:33 To @.> Cc @.> , @.> Subject Re: [swz30/MPRNet] training problem (Issue #138) image.png (view on web) 老哥，那我这个是啥问题？我按照readme来，只是改了120行和121行，因为报错说不是一个张量，后面改成了这样 image.png (view on web) 可以PSNR上不去，而且训练的很快，我数据集有2000张图的，我怀疑是不是没有正确读图？ —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: @.>

我在3090上跑的，数据集是kaggle上的夜间雨数据集， 老哥可以分享下github你看的那个吗？我翻了两遍都不知道咋改

[Feecuin]

老哥，我在yml那儿把save_images设置为True，可是我跑完并没有看到图片被保存，这是不是我读图那儿就有问题了

[Feecuin]

老哥可以看看你读数据那块的代码吗？看看你的图是按照咋样存放的，我是这样存放的没问题吧？

[Feecuin]

他readme给了可以那个google云盘的，我下不了，看不到他数据集是按照什么格式存放的

[Feecuin]

代码可以运行，读取数据应该是没有问题的 @. … ---- Replied Message ---- From @.> Date 4/1/2024 14:49 To @.> Cc @.> , @.> Subject Re: [swz30/MPRNet] training problem (Issue #138) 老哥可以看看你读数据那块的代码吗？看看你的图是按照咋样存放的，我是这样存放的没问题吧？ image.png (view on web) —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you are subscribed to this thread.Message ID: @.>

我感觉是没读进去的，因为每五轮保存图片都是空的，而且我2000张图，一个epoch竟然几秒钟结束

[Feecuin]

老哥说啥？方便加个联系方式请教下不？

[ak-karimzai]

老哥，那我这个是啥问题？我按照readme来，只是改了120行和121行，因为报错说不是一个张量，后面改成了这样 可以PSNR上不去，而且训练的很快，我数据集有2000张图的，我怀疑是不是没有正确读图？

@Feecuin Hi, can you share please the details of training?

[Feecuin]

老哥，那我这个是啥问题？我按照readme来，只是改了120行和121行，因为报错说不是一个张量，后面改成了这样 可以PSNR上不去，而且训练的很快，我数据集有2000张图的，我怀疑是不是没有正确读图？

@Feecuin Hi, can you share please the details of training?

ok, but I still haven't succeeded in training